Metraipolis › archive › dc_eea4998d3983

TAM: Task-Aware Memory Distillation for Efficient Spatiotemporal Prediction — arXiv:2610.11617

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.11617
License (as archived)
CC0-1.0
Archived
2026-10-09T06:00:21.912Z
SHA-256 of the archived text
92ed2f8aedcd6d0b924b8f83b909a94b54096df5c23471396deca874dd5f0b72
Stable URL of this page
https://metraipolis.com/documento/dc_eea4998d3983

How to cite

TAM: Task-Aware Memory Distillation for Efficient Spatiotemporal Prediction — arXiv:2610.11617 — arXiv (en), https://arxiv.org/abs/2610.11617 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eea4998d3983 (Metraipolis archive dc_eea4998d3983, sha256 92ed2f8aedcd6d0b924b8f83b909a94b54096df5c23471396deca874dd5f0b72)
BibTeX
@misc{dceea4998d3983,
  title = {TAM: Task-Aware Memory Distillation for Efficient Spatiotemporal Prediction — arXiv:2610.11617},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.11617},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eea4998d3983; sha256 92ed2f8aedcd6d0b924b8f83b909a94b54096df5c23471396deca874dd5f0b72},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.11617. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Knowledge distillation enables efficient spatiotemporal prediction by transferring knowledge from an accurate teacher to a compact student. However, matching outputs or features independently for each sample leaves cross-sample predictive structure underused. Exploiting this structure requires representations and historical references that reflect the dynamics of each task. We propose TAM, a Task-Aware Memory Distillation framework that organizes a frozen teacher's knowledge into a bounded, retrievable history. Memory entries encode latent features, forecast changes, or flow residuals, while task-specific selection rules identify relevant historical references. The student either matches the teacher's similarity distribution over shared references or regresses observation-conditioned residual prototypes. These objectives complement supervised prediction and conventional distillation. The teacher, memory, and auxiliary adapters are used only during training, leaving student inference unchanged. We evaluate TAM on video prediction, weather forecasting, and traffic flow prediction across multiple teacher-student configurations. Averaged over four paired runs, adding TAM improves SSIM on all six video datasets and reduces MSE on five relative to the corresponding KD baselines. Mean paired MSE reductions reach 1.86% on KittiCaltech, 1.93% on WeatherBench with a gSTA teacher, and 1.01% on TaxiBJ. These results demonstrate the utility of historical teacher supervision across distinct forecasting tasks without additional student inference cost.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.11617 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-09T05:47:01.800Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).